Emotion-induced blindness reflects competition at early and lateprocessing stages: An ERP study

Briana L. Kennedy & Jennifer Rawding & Steven B.Most &James E. Hoffman

# Psychonomic Society, Inc. 2014

Abstract Emotion-induced blindness (EIB) refers to im-paired awareness of items appearing soon after an irrelevant,emotionally arousing stimulus. Superficially, EIB appears tobe similar to the attentional blink (AB), a failure to report atarget that closely follows another relevant target. Previousstudies of AB using event-related potentials suggest that theAB results from interference with selection (N2 component)and consolidation (P3b component) of the second target intoworking memory. The present study applied a similar analysisto EIB and, similarly, found that an irrelevant emotionaldistractor suppressed the N2 and P3b components associatedwith the following target at short lags. Emotional distractorsalso elicited a positive deflection that appeared to be similar tothe PD component, which has been associated with attempts tosuppress salient, irrelevant distractors (Kiss, Grubert,Petersen, & Eimer, 2012; Sawaki, Geng, & Luck, 2012;Sawaki & Luck, 2010). These results suggest that irrelevantemotional pictures gain access to working memory, evenwhen observers are attempting to ignore them and, like theAB, prevent access of a closely following target.

Keywords Emotion . Attention . ERP

Introduction

Emotionally arousing stimuli can impair visual awareness ofitems presented a short time later—an effect known asemotion-induced blindness (EIB; Most, Chun, Widders, &Zald, 2005; Wang, Kennedy, & Most, 2012). In a typicalEIB task, participants search for a single target picture em-bedded in a series of rapidly presented pictures. When thetarget is preceded at short intervals by a task-irrelevant emo-tional distractor, accuracy in reporting the target is impaired.What mechanisms underlie this impairment? Are the targetssuppressed early in perceptual processing? Or do they receiveextensive perceptual and semantic processing, failing to reachawareness due to later bottlenecks such as consolidation intoworking memory (WM)?

At least superficially, EIB is reminiscent of the attentionalblink (AB), a phenomenon characterized by impaired aware-ness of the second of two targets (T2) embedded in a rapidstream of stimuli when it appears shortly after the first target(T1). One influential theory of the AB postulates that visualawareness in this context emerges via two stages of process-ing: an initial high capacity stage responsible for producing aperceptual and semantic representation of the object and asecond, capacity-limited, serial stage in which this represen-tation is consolidated into WM (Chun & Potter, 1995). If thesecond target appears while the first target is still being con-solidated, it must queue for stage 2 resources and remainsvulnerable to decay or masking by subsequent items in thestream (Dell’Acqua, Jolicoeur, Luria, & Pluchino, 2009; Dux,Asplund, & Marois, 2009; but see also Olivers, van derStigchel, & Hulleman, 2007). Variants of this theory fall underthe rubric of a central interference theory to underscore theclaim that blinked items are fully processed through percep-tual and semantic levels and are missed because of interfer-ence at the level of WM consolidation (Vogel, Luck, &Shapiro, 1998).

B. L. Kennedy (*) : S. B. MostSchool of Psychology, The University of New South Wales, Sydney,NSW 2052, Australiae-mail: b.kennedy@student.unsw.edu.au

S. B. Moste-mail: s.most@unsw.edu.au

J. RawdingPrinceton University, Princeton, NJ 08544, USA

S. B. Most : J. E. HoffmanDepartment of Psychology, University of Delaware, Newark,DE 19716-2577, USA

J. E. Hoffmane-mail: hoffman@psych.udel.edu

Cogn Affect Behav NeurosciDOI 10.3758/s13415-014-0303-x

Neuroimaging studies have provided support for the centralinterference theory of the AB. In particular, methods with hightemporal resolution, such as the event-related brain potential(ERP) and magnetoencephalography (MEG), are capable ofproviding insight into the rapid mental processes involved inresponding to rapid serial visual presentations of objects inAB and EIB studies. Vogel et al. (1998) examined severalERP components elicited by the second target in an AB taskand found that the P3b, which they suggested reflects consol-idation of information intoWM (e.g., Vogel & Luck, 2002), issuppressed during the blink. In contrast, several ERP compo-nents were unaffected by the blink, including the P1 and N1components, reflecting early perceptual and attentional pro-cesses, as well as the N400, reflecting activation of semanticinformation (e.g., Vogel et al., 1998; see also Luck,Woodman,& Vogel, 2000; Vogel & Luck, 2002). These results areconsistent with a central locus of interference, since theysuggest that processes involved in identification and semanticactivation—thought to be the province of stage 1—remainintact during the AB, whereas stage 2 processes involvingconsolidation of information into WM, reflected by the P3bcomponent, are suppressed.

The notion that the AB stems from competition for alimited-capacity system such as WM consolidation yieldsthe prediction that there should be a trade-off in the amplitudeof the P3bs associated with T1 and T2. In other words, a largeP3b elicited by T1 (reflecting access to the limited-capacitysystem) should be accompanied by a small P3b to T2,reflecting a reduction in the degree to which T2 accesses thissystem. Furthermore, this trade-off should be mirrored by asimilar pattern in the accuracy of reporting T1 and T2. Con-sistent with such predictions, this pattern of results emergedfrom an AB study that employed MEG (Shapiro, Schmitz,Martens, Hommel, & Schnitzler, 2006). When the targetswere separated by a large, 600-ms interval, participants wereable to identify both targets, and both targets elicited a largeM300 (the MEG equivalent of the P3b). However, when T2appeared 200 ms after T1, detection of T2 was reduced, andfailures to report T2 were accompanied by both an enhancedM300 for T1 and a reduced M300 for T2. However, when T2was correctly reported, the M300 for T1 was small, and theM300 for T2 was enhanced. This evidence is consistent withthe claim that the P3b/M300 reflects a limited capacity re-source involved in the AB and underscores the potential roleof WM consolidation processes in the AB.

Additional evidence suggests that a component that may berelated to selection of the second target is also suppressed inthe blink. Sergent, Baillet, and Deheane (2005) reported thatboth the P3b and the earlier, posterior N2 components (latencyof 270 ms) associated with T2 were suppressed during theblink, a finding that has been replicated several times(Kranczioch, Debener, Maye, & Engel, 2007; Reiss,Hoffman, Heyward, Doran, & Most, 2008). The posterior

N2 observed in AB might be related to the N2pc component,which indexes selection of an object in the presence ofdistractors (Woodman, Arita, & Luck, 2009). Sergent et al.noted that the N2–P3b sequence of components may reflectsequential stages in the operation of a single, limited-capacitysystem. This system might be responsible for selecting arelevant object from distractors and consolidating it into WM.

The phenomenological similarity between EIB and the ABraises the possibility that wemight observe the same pattern ofERP components during EIB as occur during the AB. Theemotional distractor images, although irrelevant, could cap-ture attention automatically, thereby gaining preferential ac-cess to WM consolidation while blocking the closely follow-ing relevant target from gaining similar access. In other words,the irrelevant, negative distractor picture should elicit an N2,reflecting its involuntary attentional selection, followed by aP3b reflecting its consolidation into WM. At short lags, whenawareness of the target is suppressed by the preceding nega-tive picture, the amplitudes of these same components that areelicited by the target should be reduced. Furthermore, both theN2 and P3b components elicited by the negative distractorshould be larger on trials associated with errors in reportingthe target, as compared with correct trials. These results wouldprovide an explanation for EIB in terms of involuntary captureof the same limited-capacity bottleneck that is implicated instudies of AB.

In summary, if EIB reflects the same trade-off in access to alimited-capacity bottleneck as that observed in studies of theAB, we should find that the irrelevant emotional distractor inthe EIB paradigm elicits a P3b component and that it is largeron error trials than on correct trials. A similar pattern of resultsshould be observed for the N2 component, since it appears tobe a reflection of an earlier stage in this same limited-capacitybottleneck. The present study examined this prediction.

Method

Participants

Twenty-six participants (9 women, 17 men; mean age,21.8 years; age range, 19–27) were recruited through a clas-sified ad and were compensated at a rate of $10 per hour. Sixparticipants were eliminated for the following reasons. Oneparticipant had overall performance accuracy near chance inall conditions. One participant did not show EIB, performingabove 90 % accuracy in all conditions. Four participants wereeliminated for having recordings with greater than 20 % ofsegments rejected, due to a combination of noisy channels,blinks, or eye movements. The results are based on the re-maining 20 participants. All participants were right-handedand reported normal or corrected-to-normal vision. Each par-ticipant provided informed consent, and the experiments were

Cogn Affect Behav Neurosci

approved by the University of Delaware Human SubjectsReview Board.

Materials and procedure

Participants completed the experiment in an electricallyshielded room. They viewed stimuli on a 17-in. Dell CRTmonitor (1,024 × 768 pixel resolution) with a refresh rate of100 Hz. Stimuli were 320 × 240 pixel color photographspresented on a gray background. Each picture subtended8.3° × 6.5° of visual angle at a viewing distance of 70 cm(participants used a chinrest). Stimulus presentation was con-trolled by software written inMATLABwith the Psychophys-ics Toolbox extensions (Brainard, 1997; Pelli, 1997).

The experiment consisted of 650 trials divided into fiveblocks. On each trial, participants viewed a sequence of 17color pictures, with each image replacing the previous oneevery 100 ms. Pictures appeared in the center of the screenagainst a gray background. The participants’ task was tosearch for a single target, which was a rotated image (90° leftor right) among a series of upright images (see Fig. 1). Fiftytrials contained no target or distractor and were not included inthe analyses; the remaining 600 trials were divided evenly in a2 (“lag”) × 3 (distractor type) design. The target was presentedeither two (lag 2) or eight (lag 8) positions after a distractorpicture, which randomly appeared as the fourth or sixth pic-ture in the stream. Distractor pictures were one of three types:emotionally arousing negative, neutral, or “baseline” (anotherscene picture randomly selected from the same set as the“filler” items in the stream). Participants indicated the direc-tion in which the target picture was rotated (left or right) and

rated their confidence on a 3-point scale (very sure, somewhatsure, or guess).

One hundred twenty-eight images served as targets (64landscape and architectural images rotated 90° both clockwiseand counterclockwise). Filler images came from a bank of 252upright landscape and architectural photographs. There were56 negative distractor pictures (depictions of medical trauma,threatening animals, and violence) and 56 neutral distractorpictures (depictions of people or animals with no obviousemotional content). In the baseline condition, the item occu-pying what would have been the serial position of thedistractor was drawn from the same bank of landscape/architectural images as the filler items. Negative and neutralimages were primarily drawn from the International AffectivePicture System (Lang, Bradley, & Cuthbert, 2001) on the basisof ratings of valence and arousal and were supplemented withsimilar images from publicly available sources.

At the start of each trial, participants viewed a fixationcross in the center of the screen and clicked the mouse whenthey were ready to view the stream of stimuli. They wereinstructed not to move or blink their eyes during the presen-tation of the images. After the stream of stimuli, a blank screenappeared for 1 s, followed by a response screen depicting sixbuttons. The six buttons were arranged into two groups ofthree, which corresponded to rotation direction (left vs. right)and confidence (very sure, somewhat sure, or guess). Catego-rization of response accuracy incorporated the confidencejudgments as a way to eliminate correct guesses. Responseswere categorized as correct if (1) they matched the orientationof the target picture and (2) the corresponding confidencerating was either very sure or somewhat sure. Responses werecategorized as incorrect if participants reported the wrongorientation or reported that they were guessing.

Before starting the experiment, participants were explicitlytold that images of people and animals would appear in theexperiment, that they would sometimes be unpleasant, andthat they would never be the rotated target. They were alsoshown examples of emotional and neutral images to ensurethat they were comfortable completing the experiment andwere reminded several times that they could withdraw fromthe experiment at any time. They then engaged in a short 16-trial practice session, with RSVP rates starting at 5 images persecond and increasing to the experiment presentation rate of10 images per second. The practice session did not includedistractors. Participants were debriefed at the end of theexperiment.

Electrophysiological recording and data analysis

The electroencephalogram (EEG) was recorded with an Elec-trical Geodesics Inc. system (EGI; Eugene, OR) using a 129-channel Hydrocel Sensor Net. Individual electrode imped-ances were kept below 50–75 kΩ, as recommended by the

Fig. 1 a Types of distractors and b schematic design of the study. See thetext for details

Cogn Affect Behav Neurosci

manufacturer. The data were referenced online to the vertex,band-pass filtered from 0.01 to 80 Hz, and digitized at 200 Hz.Subsequent processing was performed offline using EGI NetStation 4.1.2 software. The data were low-pass filtered with acutoff of 40 Hz and then segmented using an epoch that began200 ms prior to the onset of the distractor picture and ended1,200 ms after onset. Channels were marked as bad if themaximum voltage range exceeded 100 μV. Individual seg-ments were rejected if more than 10 channels were marked asbad. Trials containing blinks (threshold = 100 μV) or eyemovements (threshold = 70 μV) were rejected. For the re-maining segments, bad channels were replaced by interpolat-ing from surrounding channels. Finally, the segments wereaveraged, rereferenced to the average reference, and baselinecorrected using the 200-ms prestimulus interval.

Data analysis procedures

Difference waves were used to isolate ERP components elic-ited by distractor and target pictures from the periodic ERPactivity generated by the sequence of pictures (see Vogel et al.,1998). This is particularly important in the presentexperiment, which used a rapid sequence of pictures. Eachbackground picture produces its own set of sensorycomponents every 100 ms, and effects of the various criticalevents, such as the presentation of the distractor and targetpictures, produce their own ERPs superimposed on thebackground activity. As Vogel et al. (1998) demonstrated,subtracting the ERP in one condition from another can isolatethe ERPs associated with the critical events. For example,ERP components elicited by the negative distractor picturecan be isolated by subtracting the baseline lag 8 conditionfrom the negative lag 8 condition. In the first 1,000 ms of therecording epoch, these conditions differ only in the presenceor absence of the distractor, so the subtraction will isolate theERP components due to the presence of a distractor. Similarly,components elicited by the target picture can be isolated bysubtracting the lag 8 condition from the corresponding lag 2condition for each type of distractor. For example, the com-ponents elicited by the target picture when it is preceded by anegative distractor can be isolated by subtracting the negativelag 8 condition from the negative lag 2 condition. Both con-ditions have a negative distractor in common, so the associat-ed components will subtract out. The two conditions differonly in the presence or absence of a target, so target compo-nents will be isolated in this subtraction.

The sensor locations corresponding to the two componentsof interest in this study (the N2/EPN and the P3b) were chosenusing a method recommended by Keil et al. (2014) to mini-mize the inflation of type 1 error that would accompanychoosing sensors on the basis of the largest signal observedin the entire set of sensors. We first averaged over the threedistractor conditions separately for the lag 2 and lag 8

conditions. We then computed the lag2 − lag8 subtraction,which should reveal the P3b and N2 components elicited bythe target. A second subtraction used the average acrossdistractor conditions at lag 8 but subtracted the baseline lag8 curve to isolate the N2 elicited by the distractors. As isdetailed below, we also observed a positivity following thenegative and neutral distractor pictures that had the same scalpdistribution as the N2 so we used these sensors to measure itsamplitude. Clusters of three to six spatially contiguous sensorsthat showed themaximum amplitude in the relevant differencewave were averaged together. These difference curves werealso used to specify time windows for measuring averageamplitude. The beginning and end points of the averagingwindow were chosen corresponding to the points at whichthe amplitude was half of the peak value (Picton et al., 2000).

The sensors that we chose for the N2/EPN were centeredon PO7 and PO8 in the left and right hemispheres, respective-ly, while the P3b was centered on a location midway betweenPz and CPz. These locations are in good agreement withprevious research on these two components (e.g., Schupp,Junghöfer, Weike, & Hamm, 2003a). Amplitude measureswere analyzed with a repeated measures analysis of variance(ANOVA), which employed Greenhouse–Geisser correctionsfor violations of sphericity. Significant main effects involvingthree levels of a factor (e.g., negative, neutral, and baselinedistractors) were followed up with least significant differencetests between each pair of means. This procedure does notinvolve a correction to the alpha level for multiple compari-sons, because there is no inflation of family-wise error ratesfor the special case of three conditions, as long as post hoctests are preceded by a significant main effect (Cardinal &Aitken, 2006).

As is described in the Results section, determining whetherthe distractor picture produced a P3b component was compli-cated by the finding that the negative and neutral distractorpictures elicited a prominent positivity over posterior sensorsin a time window similar to what one would expect for a P3bto the distractor picture. This posterior positivity had a scalpdistribution that was different from the typical topography ofthe P3b component, but it appeared to partially overlap withsensors in the P3b region. In order to objectively separate thiscomponent from the P3b, we analyzed the data using a two-step principal components analysis (PCA) that employed theERP PCA (EP) Toolkit version 2.34 for MATLAB (Dien,2010). Analyses were based on the covariance matrix withKaiser normalization. Promax rotations were used for bothtemporal and spatial steps, thus allowing factors to be corre-lated. The “parallels” test (Dien, 2010) was used to determinethe number of factors to be retained in each step. The spatialstep was performed first in order to capitalize on the promi-nent P3b component elicited by targets at lags 2 which shouldyield a clear spatial P3b factor. A subsequent temporal PCAwas conducted using the spatial factors from the first step as

Cogn Affect Behav Neurosci

“virtual electrodes” (Spencer, Dien, & Donchin, 2001). Thetemporal factors for the P3b spatial factor should include atime window for the target. Of primary interest was whetherwe would also obtain an earlier time window corresponding toa P3b elicited by the emotional distractor.

In the first step, a spatial PCAwas performed on data fromthe 129 sensors using average waveforms for each subjectconsisting of the three distractor type (negative, neutral, andbaseline) at lag 2. Each waveform included 280 time pointscovering the interval from 200 ms before the appearance ofthe distractor to 1,200 ms following it. On the basis of theparallels test (Dien, 2010), we retained 10 spatial factors. Inthe second step, a temporal PCA was performed on the timepoints using the combinations of distractor type (negative,neutral, and baseline) × 10 “virtual spatial electrodes” (thespatial factors derived in the first step). Eight temporal factorswere retained, yielding 80 factor combinations (10 spatial × 8temporal factors). We inspected this set of factors to locate theP3b spatial component, which was defined as having a mid-line positive maximum over posterior scalp locations (cen-tered on Pz) and having a large amplitude in response totargets in the baseline condition at lag 2. This approachresulted in a clear P3b spatial factor, allowing us to examinethe eight temporal windows to determine whether a P3b waselicited by the distractor pictures.

Results

Behavioral results

Figure 2 shows percent accuracy (corrected for guessing; seethe Method section) in discriminating the orientation of thetarget when it was preceded by different distractor types(negative, neutral, and baseline) at two different lags. A two-factor ANOVA (distractor type: negative, neutral, baseline ×lag 2 vs. 8) revealed significant effects of distractor type, F(2,38) =72.37, p < .001, and lag, F(1, 19) = 133.69, p < .001, andtheir interaction, F(2, 38) = 87.73, p < .001. A separate

analysis of the lag 2 data revealed a significant effect of valence,F(2, 38) = 108.84, p < .001. Post hoc tests showed that all pairsof distractor types were significantly different from each other(all ps < .001). This pattern of results is similar to that observedin previous studies of EIB (e.g., Most et al., 2005).

ERP results

N2

The N2 component elicited by the distractor pictures wasisolated by subtracting the lag 8 baseline condition from thenegative and neutral lag 8 conditions. In Fig. 3, the resultingsubtraction ERPs are shown in the form of a topographic map(negative distractor, latency = 225 ms) and waveforms. Thetopographic map shows that the distractor N2 is broadlydistributed over posterior areas in the left and right hemi-spheres, with a slight maximum amplitude over the righthemisphere. This topography is consistent with previous re-ports of the N2 component to emotional stimuli (the N2elicited by emotional pictures is often referred to as an earlyposterior negativity, or EPN; e.g., Schupp et al., 2003a). TheN2 peaks 225 ms after the onset of the distractor picture and islarger for the negative than for the neutral distractor. Wequantified this effect by averaging over three contiguoussensors in each hemisphere in the vicinity of T5 and T6 (64,68, and 69 in the left hemisphere; 89, 94, and 95, in the right).The time window (185–275 ms after distractor onset)corresponded to the half amplitude points on either side ofthe peak (Picton et al., 2000). Although average voltage in thewindow is the preferred measure of component amplitude, itwould produce a distorted value for the neutral distractor N2as it is of shorter duration than the negative distractor N2 andreverses polarity before the end of the measurement window.We circumvented this problem by using the peak negativity inthe window as an amplitude measure for this comparison only.A two-factor ANOVA (hemisphere × distractor type: negativeand neutral) on the peak values revealed that the N2was largerfor a negative distractor than for a neutral one, F(1, 19) =28.28, p < .001. Neither the main effect of hemisphere, F(1,19) = 2.47, p = .10, nor its interaction with distractor type, F <1, was significant. In addition, single-sample t-tests showedthat the peak amplitude was significantly different from zerofor both the negative distractor, t(19) = 7.24, p < .001, andneutral distractor, t(19) = 2.90, p = .009, indicating that bothN2s were larger than the N2 observed in the baselinecondition.

The N2 component elicited by the target was isolated bysubtracting the corresponding lag 8 condition from each lag 2condition. For example, the target N2 in the negative distractorcondition was obtained by subtracting the negative lag 8condition from the negative lag 2 condition. This subtractionisolates the target effect, because the only difference betweenFig. 2 Behavioral results. See the text for details

Cogn Affect Behav Neurosci

these conditions in the first 800 ms postdistractor is the pres-ence of a target picture in the lag 2 condition. A similarsubtraction was used to isolate the target N2 in the neutralcondition (lag 2 neutral minus lag 8 neutral) and in thebaseline condition (lag 2 baseline minus lag 8 baseline).

Figure 4 shows the resulting topographic maps and wave-forms. The scalp distribution of the target N2 is similar to thatof the distractor N2, with a broad posterior distribution and aslight maximum over the right hemisphere. The peak negativ-ity occurred at 250 ms following the onset of the target. Theamplitude of the target N2 was largest when it was precededby a baseline distractor and smallest when it was preceded bya negative distractor. Amplitude was intermediate in the caseof a neutral distractor. To quantify these effects, we used thesame six sensors (three in each hemisphere) used above for thedistractor N2 and a time window of 205–385 ms followingtarget onset. An ANOVA using the factors of hemisphere anddistractor type showed no main effect of hemisphere, F < 1,and a significant main effect of distractor type,F(2, 38) = 9.51,p < .001. The interaction was not significant, F < 1. Post hocpaired comparisons revealed that the target N2 was signifi-cantly larger when the target was preceded by a baselinedistractor, as compared with either a negative (p < .001) or aneutral (p = .04) distractor. The negative and neutral distractorconditions were also significantly different (p = .04).

Target P3b

The P3b component elicited by the target picture at lag 2 foreach condition was isolated by subtracting the lag 8 waveformfrom its corresponding lag 2 waveform for each of the threedistractor conditions. These waveforms are shown in Fig. 5. Onthe basis of a combination of prior research (e.g., Polich, 2012),a PCA we conducted (see below), and the observed scalp

topography, we quantified the P3b as the average amplitudefor sensors 54, 55, 79, 78, 61, and 62, which is located betweenCz and Pz in the 10-20 system (white dots in Fig. 5). Themeasurement window extended from 435 to 795 ms relative totarget onset (635–995 ms relative to distractor onset). A one-way repeated measures ANOVA revealed a main effect ofdistractor type, F(2, 38) = 16.11, p < .001. Pairwise compari-sons showed that the negative distractor condition produced a

Fig. 3 Distractor N2 topographyand waveforms. The topographicmap displays the negative minusbaseline distractor conditionswhen the target appeared at lag 8(at the peak latency of 225 mspost-distractor-onset). Waveformswere averaged across contiguouselectrode sites (represented inwhite) and show the negative-baseline and neutral-baselinedifference curves when the targetappeared at lag 8. The grayrectangle shows the measurementwindow (185–275 ms post-distractor-onset)

Fig. 4 N2 topography and waveforms elicited by the target pictureappearing at 200 ms. The topographic map displays the negative minusbaseline lag 2 conditions (at the peak latency of 250ms post-target-onset).Waveforms were averaged across contiguous electrode sites (representedin white) and show the negative lag 2 − negative lag 8, neutral lag 2 −neutral lag 8, and baseline lag 2 − baseline lag 8 subtraction waveforms.The gray rectangle shows the measurement window (205–385 ms post-target-onset)

Cogn Affect Behav Neurosci

smaller target P3b than did both the neutral (p < .001) andbaseline (p < .001) conditions. The neutral and baseline condi-tions were not significantly different (p = .42). These results aregenerally consistent with the behavioral accuracy data, whichshowed a large blink for the negative distractor and a muchsmaller one for the neutral condition. It might be that thedifference in accuracy between the neutral and baseline condi-tions was simply too small to reliably affect P3b amplitude.

Distractor P3b

In order to determine whether the distractor pictures elicited aP3b, we subtracted the lag 8 baseline condition from the lag 8negative and lag 8 neutral conditions. The resulting

subtraction curves are shown in Fig. 6. The waveforms rep-resent the average of the same six sensors that were used toquantify the target P3b. For illustration purposes, this figurealso contains the waveform for the target P3b elicited in thebaseline condition (lag 2 baselineminus lag 8 baseline), whichdid not have a P3b in the time interval corresponding to thedistractor. To quantify P3b amplitude, we used a measurementwindow extending from 400 to 550 ms following distractoronset. A one-way repeated measures ANOVA revealed asignificant effect of distractor type, F(2, 38) =14.36, p <.001. Pairwise comparisons showed that the negativedistractor condition produced a larger P3b than did both theneutral (p < .001) and baseline (p = .012) conditions. Theneutral and baseline conditions were not significantly different

Fig. 5 Scale topographies anddifference curves for the P3b tothe target picture appearing at200 ms. The topographic mapdisplays the negative minusbaseline lag 2 conditions at thelatency of 520 ms post-target-onset (720 ms post-distractor-onset). Waveforms were averagedacross contiguous electrode sites(represented in white) and showthe negative lag 2 − negative lag8, neutral lag 2 − neutral lag 8,and baseline lag 2 − baseline lag 8subtraction waveforms. The grayrectangle shows the measurementwindow (435–795 ms post-target-onset)

Fig. 6 Distractor P3btopographies and differencecurves. The topographic mapdisplays the negative minusbaseline lag 8 conditions (at thelatency of 485 ms post-distractor-onset). The negative and neutralwaveforms represent distractorswith no target presented until lag8 (using the negative-baseline andneutral-baseline differencewaves). The baseline conditionserves as a comparison and showsthe P3b for lag 2 targets thatappear after baseline distractors(lag 2 baseline − lag 8 baselinedifference wave). The grayrectangle shows the measurementwindow (400–550 ms post-distractor-onset). See the text fordetails

Cogn Affect Behav Neurosci

(p = .053), and no P3b was apparent for the neutral distractor.We revisit this issue in the PCA section below.

Posterior positivity

The N2 to the distractor was followed by a positivity with asimilar scalp distribution, shown in Fig. 7 for the negative andneutral distractor conditions. This component was isolated bysubtracting the lag 8 baseline condition from the lag 8 negativeand lag 8 neutral conditions. None of these contained a targetin the first 800 ms of the recording epoch, so any resultingactivity should be attributable to the distractor pictures. Thesubtraction topography shows a bilateral, posterior positivityover occipital temporal areas that is slightly larger over theright hemisphere than over the left. We will refer to this as theposterior positivity component, although we speculate in thediscussion that it might be identical to the PD component thatis elicited by salient, task-irrelevant singletons in visual searchtasks (Kiss, Grubert, Petersen, & Eimer, 2012; Sawaki et al.,2012; Sawaki & Luck, 2010). We quantified the averageamplitude of this component using a measurement windowof 405–615 ms after distractor onset. The sensors used toassess its amplitude were 96, 100, and 101 in the right hemi-sphere (close to T6 and O2 in the 10–20 system) and 64, 68,and 69 in the left hemisphere. Notably, these sensors aresimilar to those used to measure the earlier N2 component,and the peak latency of this component (470ms) is close to thelatency of the N2 to lag 2 baseline targets (265 ms posttarget

and 465 ms postdistractor). This component can also be seenin the topographic map in Fig. 6, where it overlaps with theP3b elicited by the negative distractor. An ANOVA revealedthat the posterior positivity was larger for negative than forneutral distractors, F(1, 19) = 16.68, p = .001. There was nomain effect of hemisphere, F(1, 19) = 1.31, p = .268, or of theinteraction of hemisphere and distractor type, F(1, 19) = 4.19,p = .055. However, the partial overlap in time and sensorlocations between the posterior positivity and the P3b elicitedby the negative distractor makes it difficult to know howmuchour measurement of the posterior positivity is confounded byeffects of the distractor P3b. We addressed these issues byseparating the components using a PCA.

Principal components analysis

In order to separate the P3b components elicited by the neg-ative distractor and the following target at lag 2 and to ensurethat the posterior positivity is a separate component from thedistractor P3b, we conducted a two-step, spatial-temporalPCA (Dien, 2010; see the Method section) on the 1,200-msinterval following onset of the distractor picture.

The first step was a spatial PCA. This should isolate aspatial P3b component due to the presence of a robust P3bto the lag 2 target in the baseline condition. Applying atemporal PCA to this spatial factor should produce two timewindows corresponding to P3bs elicited by the distractor andthe lag 2 target. We retained 10 spatial factors from the firststep and 8 temporal factors from the second step (see theMethod section for details). Figure 8 shows the scalp topog-raphy for spatial factor 2, which corresponds well with theknown distribution of the centro-parietal P3b. It is also con-sistent with the topography of the target P3b shown in Fig. 5.Temporal factors one and three (S2T1 and S2T3) for thisspatial component peak at 575 ms following the target and485 ms following the distractor. Figure 8 plots these factorsfor the baseline lag 2 and negative lag 8 conditions, respec-tively. In order to facilitate comparison with the earlier peakanalysis, we quantified the PCA for the lag 2 target (S2T1) interms of the difference between lags 2 and 8.These differencescores were 2.01, 4.03, and 4.42 μV for negative, neutral, andbaseline conditions, respectively. An ANOVA on these datarevealed a significant effect of distractor type, F(2, 38) =22.17, p < .001. Post hoc comparisons showed that the neg-ative distractor condition resulted in a smaller target P3b thandid both the neutral (p < .001) and baseline conditions (p <.001), which did not differ (p = .28). These results are inagreement with the results of the corresponding peak analysispresented earlier in showing that the P3b to the target wassuppressed following a negative distractor.

A similar analysis was conducted on factor scores forS2T3, which appears to be a P3b elicited by the distractorpicture. A two-factor (distractor type × lag) ANOVA

Fig. 7 Topographic map and waveforms for the posterior positivitycomponent. The topographic map displays the neutral minus baselinedistractor conditions (at the peak latency of 470 ms post-distractor-onset)in the lag 8 condition. Waveforms were averaged across contiguouselectrode sites (represented in white and indicated with arrows) and showthe negative − baseline and neutral − baseline difference curves when thetarget appeared at lag 8. Electrode sites for the P3b are also shown forreference. Waveform data have been averaged over hemispheres. Thegray rectangle shows the measurement window (265–465 ms post-distractor-onset)

Cogn Affect Behav Neurosci

conducted on the factor scores revealed no main effect orinteraction involving lag (F < 1 and F(2, 38) = 1.16, p = .325,respectively). There was a main effect of distractor type, withmeans of 1.51 V, −0.68 V, and −0.21 V for the negative, neutral,and baseline conditions, respectively. This resulted in a maineffect of distractor type, F(2, 38) = 23.15, p < .001. Post hoccomparisons showed that the distractor P3b was larger in thenegative condition than both the neutral (p < .001) and thebaseline conditions (p = .001). The neutral and baseline condi-tions were not significantly different from each other (p = .094).These results provide a confirmation of the results obtained withthe amplitude measures of the distractor P3b presented earlierand indicate that the negative distractor elicited a P3b compo-nent even though it was irrelevant to the participant’s task.

The PCA also revealed factors corresponding to the poste-rior positivity component. Separate factors were found for theleft (S6T2, latency = 465 ms) and right (S7T2, latency =465 ms) hemispheres, and topographic plots of these factorsare shown in Fig. 9, along with waveform data for the threedistractor types in the lag 8 condition. An ANOVA with thefactors of distractor type, hemisphere, and lag showed nosignificant main effect of hemisphere (F < 1) or any of itsinteractions with other variables (all ps > .15). The waveformdata, averaged over hemisphere, show that the posterior pos-itivity was largest for negative distractors, intermediate forneutral distractors, and smallest for the baseline distractors,resulting in a significant distractor effect, F(2, 38) = 71.85, p <.001. Pairwise comparisons revealed that all three distractorconditions were significantly different from each other (all ps< .001).

The important contribution of the PCA is the confirmationthat the negative distractor picture elicited a P3b componentthat was separate from a more posterior positivity that

occurred in the same time window. We discuss the import ofthese findings in the Discussion section. But first, we exploredwhich of these components were related to accuracy inreporting the target.

Incorrect versus correct trials

By examining which ERP components were affected bywhether participants were correct or not on the target discrim-ination, one can identify those components that are related tothe “blink” produced by the negative pictures. We were

Fig. 9 Posterior positivity topography for spatial factors 6 (left hemi-sphere, LH) and 7 (right hemisphere, RH) from the principal componentanalysis (PCA). Topographic maps display these factors at peak latencies(505 and 465 ms post-distractor, respectively). Waveforms represent theaverage of the LH and RH spatial factors at temporal factor 2 in the lag 8conditions for each distractor type

Fig. 8 P3b topography for spatialfactor 2 from the PCA. Thetopographic map displays at thelatency of 575 ms post-target-onset. The two waveformsdepicted represent the P3b elicitedby the negative distractor (lag 8)and a target presented at lag 2after a baseline distractor

Cogn Affect Behav Neurosci

particularly interested in whether any of the three componentsexamined above—the N2, P3b, or posterior positivity—wereassociated with errors in reporting the target. To evaluate this,we examined ERPs for correct and incorrect trials in the lag 2negative distractor condition, since this was the only conditionin which there were sufficient errors to make this analysisfeasible (across participants, the number of error trials in thiscondition ranged from 21 to 66, with a mean of 42). We usedthe same sensor clusters and time windows used above toexamine the impact of accuracy on the different components.

Distractor N2 and posterior positivity

Figure 11 shows the N2 topographic map and waveforms forcorrect and incorrect trials, as well as their difference, whichhas a bilateral distribution similar to that of the N2 (see Fig. 3);consequently, we used the same sensors to quantify it. Theincorrect − correct difference curve peaks at 290 ms, which islater than the peak we found for the distractor N2 (225 ms)when it was based on the average of correct and incorrecttrials. An examination of the separate waveforms for correctand incorrect trials makes it clear that the effect of accuracy isreflected in a longer duration of the distractor N2 on incorrecttrials. We examined the significance of this effect using awindow extending from 250 to 350 ms. A t-test showed thatthere was no significant difference between left and righthemispheres, t(19) = 0.818, p = .424, so we averaged overthem. A second t-test showed that the average amplitude wasgreater than zero, t(19) = 3.34, p = .003. These results showthat participants were more likely to make an error on the

target when the preceding negative distractor elicited a late orextended N2.

The initial negativity associated with incorrect trials wasfollowed by a positivity that appears to have the same scalptopography as the posterior positivity examined earlier (seetopographic map in Fig. 10). Therefore, we used the sameright-hemisphere sensors employed in the earlier analysis anda time window extending from 410 to 650 ms following thedistractor. A t-test showed that incorrect trials were associatedwith a larger distractor posterior positivity than were correcttrials, t(19) = 5.12, p < .001. In addition, it appears that thispositivity has an extended time course on error trials.

The prolonged duration of the distractor N2 on incorrecttrials may reflect longer engagement of attention on thesetrials. The subsequent posterior positivity may reflectdisengagement of attention, in which case its larger amplitudeand extended time course on error trials may reflect greaterdifficulty in disengaging attention. Therefore, sustained N2and posterior positivity components elicited by the distractormay index trials on which attention remained engaged on thedistractor for a prolonged period, resulting in a failure of thelag 2 target to access the attention system.

Distractor and target P3b

If errors were at least partly due to the distractor pictureoccupying WM and preventing access by the target, asappears to be the case in the AB, we would expectlarger distractor P3bs on incorrect trials than on correcttrials. Figure 11 shows the topography (at 460 ms) and

Fig. 10 Waveforms for incorrectand correct trials, as well as theirdifference (incorrect − correct).The topographic maps representthe difference scores at a latencyof 290 ms for the N2 and 535 msfor the posterior positivity.Waveforms display data from thesensors depicted in white (thesame sensors used to measure theN2 and posterior positivity inearlier analyses)

Cogn Affect Behav Neurosci

waveforms for correct and incorrect trials and theirdifference. Note that the topographic map for the targetP3b corresponds to the correct − incorrect difference inorder to show a positive P3b. To quantify the P3b, weused the same sensors and time window used above inthe section on the Distractor P3b.

The average amplitude in the distractor P3b window was0.29 μV for correct trials and 1.00 μV for incorrect trials. Thedifference between these two values was significant, t(19) =2.57, p = .02, showing that errors in reporting the target wereassociated with larger P3bs elicited by the negative distractorpicture.

We might also expect that the P3b to targets wouldbe larger on correct trials, relative to incorrect trials,because one could assume that the target was morelikely to have gained access to WM on correct trials,and this appears to be the case in the waveforms shownin Fig. 11. A t-test on the average amplitude in theinterval from 435 to 795 ms, relative to target onset(635–995 ms relative to distractor onset), confirmed thatthe target P3b was larger on correct trials (2.66 μV)than on incorrect trials (1.87 μV), t(19) = 2.74, p =.013.

These data show that, relative to correct trials, incor-rect trials are associated with a larger distractor P3b anda smaller target P3b. In other words, there is a tradingrelationship in the amplitudes of the P3bs elicited bythe distractor and target pictures, which supports thehypothesis that competition for a central resource (WMconsolidation) is at least part of the reason that irrele-vant negative distractor pictures suppress awareness forclosely following targets.

Discussion

The present experiment used ERPs to examine the mecha-nisms responsible for EIB, the ability of a task-irrelevant,negative picture to suppress awareness of a closely followingtarget picture. Given the phenomenal similarity between EIBand the AB, we were specifically interested in whether theprocesses that are thought to drive the AB might also beresponsible for EIB.

According to the central interference account of the AB, theinability to report the second of two targets presented in closetemporal proximity reflects competition at a relatively latebottleneck in target processing, such as consolidation of in-formation into WM (e.g., Chun & Potter, 1995; for reviews,see Dux & Marois, 2009; Martens & Wyble, 2010). In partic-ular, this model assumes that T1 occupies the consolidationprocess for an extended period of time, so that a closelyfollowing T2 has to queue to gain access and is vulnerableto masking by subsequent distractors. Electrophysiologicalevidence in support of this account has been provided byexperiments that used the P3b component (Kranczioch et al.,2007; Vogel et al., 1998) and the magnetic counterpart to theP300, the M300 (Shapiro et Al., 2006; for simplicity, we willrefer to both components as the P3b), as a measure of access tothe consolidation process. These experiments have shown thatthe amplitudes of the P3b associatedwith the two targets in thestream have a trading relationship: A larger amplitude P3belicited by T1 is associated with a smaller amplitude P3b forT2, and vice versa (Kranczioch et al., 2007; Shapiro et al.,2006). In addition, errors in reporting T2 are associated withlarger P3bs to T1 (Shapiro et al., 2006; Vogel et al., 1998).Both of these findings are consistent with the predictions of acentral interference account of the AB.

In the present study, we found a similar pattern of results.Negative emotional pictures elicited a P3b component eventhough they were irrelevant to the observer’s task. Important-ly, error trials were associated with larger distractor P3b com-ponents than were correct trials. These data suggest thatemotional pictures sometimes “automatically” gain access toWM consolidation, thereby blocking access by the targetpicture. Thus, central interference is a mechanism that playsa role in both EIB and AB.

The neutral distractor pictures also impaired detection ofthe following target, albeit much less than the negativedistractors (an approximately 9 % reduction in target accuracyfor neutral distractors, as compared with 28 % for negativedistractors). The neutral distractors did not appear to elicit aP3b and produced a smaller reduction in target P3b amplitude,relative to negative distractors. These results show the sametrading relationship between temporally proximal P3bs thatwas observed in the AB and suggest a similar conclusion:Larger P3bs elicited by distractor pictures reflect greater use ofa limited-capacity bottleneck mechanism that is required for

Fig. 11 Waveforms and topographic maps for the incorrect − correctdifference using the P3b sensors. Topographic plots display the incorrect− correct difference at a latency of 500 ms for the distractor and 800 msfor the target. Waveforms display data from the sensors depicted in white(the same sensors used to measure the P3b in earlier analyses)

Cogn Affect Behav Neurosci

awareness of closely following targets. Consequently, thesetargets are less likely to gain admittance to the bottleneck,resulting in smaller target P3b components. Therefore, thesame limited-capacity mechanism that is reflected in the P3bcomponent seems to play a role in target suppression in bothEIB and AB.

In addition, variations in two other components of theERP—the N2 and a posterior positivity—were related to theoccurrence of EIB. First, we consider the N2. In agreementwith earlier reports (Flaisch, Junghöfer, Bradley, Schupp, &Lang, 2008; Peyk, Schupp, Keil, Elbert, & Junghöfer, 2009;Schupp, Junghöfer, Weike, & Hamm, 2003a, 2003b), wefound that emotional pictures produced a larger N2 compo-nent, as compared with neutral and baseline pictures. Thetarget picture, when it was not preceded by a distractor, alsoelicited a prominent negative component that had a latencyand scalp distribution that was similar to the N2 elicited bydistractors. Given the similar temporal and topographic pat-terns, both target and distractor N2s appear to be produced bythe same neural generator, which suggests that the N2 elicitedby emotional pictures, known as the EPN in the emotionliterature, may not be uniquely associated with emotionalstimuli, in agreement with earlier studies (see Schupp,Flaisch, Stockburger, & Junghöfer, 2006). The significantN2 generated by the neutral distractors suggests the sameconclusion.

The N2 to the target picture was largely eliminated whenthe target was preceded by a distractor picture, and the mag-nitude of suppression was larger following negativedistractors, as compared with neutral distractors. Negativedistractors themselves produced a larger N2 than did neutraldistractors, and thus the amplitudes of the N2s for distractorsand targets exhibited a trading relationship similar to thatwhich was observed for the P3b component: Larger distractorN2s were associated with smaller target N2s. Importantly, thedistractor N2 was also related to accuracy in reporting thetarget. Trials on which the observer failed to report the targetwere associated with temporally extended N2s to thedistractor. Thus, the N2 and P3b components appear to actin concert in producing EIB.

The relatively early latency of the N2 (~230–250 ms afterstimulus onset) and its scalp distribution over the posterioroccipital-temporal cortex are consistent with this componentbeing related to perceptual and/or attentional processes. Onepossibility is that the N2 in AB and EIB paradigms is relatedto the N2pc component, which is thought to index the alloca-tion of object-based attention (Luck, 2011). Indeed, the N2observed in the present study is similar in latency and topog-raphy to the N2pc. The N2pc is normally strongly lateralizedand is measured by subtracting ipsilateral from contralateralactivity at posterior electrode sites (Luck, 2011). Our stimuliwere presented at fixation, resulting in a bilateral distributionthat precluded this subtraction approach. To the degree that

these components are the same, we speculate that the N2reflects the capture of object-based attention by a salient buttask-irrelevant picture, as well as by relevant targets. A smalltarget N2 is associated with a failure of awareness for thattarget, while a small distractorN2 reflects its failure to captureattention, thereby resulting in a reduction in EIB. These resultssuggest that EIB is at least partly due to a competition forattentional resources. When a salient negative distractor pic-ture captures attention, it elicits an N2, and this appears toreflect access to a limited-capacity bottleneck and, ultimately,consolidation into WM reflected in the P3b. In addition, theextended duration of the negative distractor N2 on error trialsmay reflect longer attentional engagement times followed byprolonged disengagement, reflected in the posterior positivitycomponent (see below). Sustained attention to the distractorwould decrease the likelihood that the target is able to gainaccess to the selection process required for entry into WM,resulting in a “blink.”

Sergent et al. (2005) reported similar findings for the ABtask. They found reduced N2 and P3b components elicited bythe second target when it was blinked (see also Krancziochet al., 2007). They noted that at lags producing a blink, the N2elicited by the second target overlapped in time with the P3belicited by the first target, suggesting that these two compo-nents (N2 and P3b) reflect a common limited-capacity processthat cannot simultaneously process two temporally proximaltarget events. They speculated that the N2 is an earlier man-ifestation of the same limited-capacity process that is reflectedin the later P3b. One possibility is that the N2 componentindexes the selection of which objects or events will be passedon to a second process, reflected in the P3b, which consolidatesinformation into WM. Thus, these two processes, and theirassociated ERP components, would generally be closelylinked. If the selection process reflected by the N2 is a bottle-neck that is limited to one object (or event; see Sheppard,Duncan, Shapiro, & Hillstrom, 2002) at a time and the durationof this process has an extended time course, it would appear tobe a candidate for playing a role in the AB and EIB.

Intriguingly, we also observed a robust posterior positivityfollowing the distractor that was larger for negative than forneutral distractors. This component, which we call the poste-rior positivity, appeared to persist for a longer time on trials onwhich the observer failed to report the target, and it had a scalptopography that was similar to that of the N2 but opposite inpolarity. This posterior positivity may be related to the Pdcomponent (Kiss et al., 2012; Luck & Hillyard, 1994; Sawaki,Geng, & Luck, 2012; Sawaki & Luck, 2010) that is elicited bytask-irrelevant distractors in attentional capture paradigms.Such distractors have been found to elicit an N2pc, followedby a Pd, which may reflect a sequence of engagement anddisengagement of attention with respect to the distractor ob-ject. Like the posterior positivity in our study, the Pd in theserecent studies had a scalp topography similar to that of the

Cogn Affect Behav Neurosci

preceding N2pc, leading Sawaki and colleagues to concludethat the “N2pc and Pd reflect related, but opposite, mecha-nisms” (Sawaki et al., 2012, p. 7). While the posterior posi-tivity component we observed strongly resembles the Pd, weemphasize that inferring that they are the same component isspeculative, and future studies need to be conducted to con-firm it.

Previous research on attention capture by emotional pic-tures has concentrated on the EPN/N2, as well as the latepositive potential or LPP, which refers to a constellation ofpositive components that appear to include the P3b and asustained positive slow wave. The LPP is larger for negativethan for neutral pictures and also appears to be related tointerference associated with irrelevant negative distractors(Weinberg & Hajcak, 2011). Most previous investigators didnot comment on whether they also observed a posteriorpositivity following negative pictures that overlaps in timewith the P3b, but it could have been overlooked because ofthis overlap. One exception is a study by Foti, Hajack, andDien (2009), who used temporal-spatial PCA to separate theLPP into the P3b, a positive slow wave, and a posteriorpositivity that peaked at 343 ms following the negativepicture, with a maximum amplitude at occipital sites.

The scalp topography of this early LPP component appearsto be similar to the posterior positivity in the present article,while its earlier onset is probably related to the slower presen-tation rates used by Foti et al. (2009; their P3b to emotionalpictures was also approximately 100 ms earlier than ours).They interpreted this component as being related to the P3b,while we are suggesting that it might reflect a different under-lying process than the P3b—namely, disengagement from theemotional picture. Consistent with the latter claim, we foundthat that the correlation between scores on the two factors inour PCA that corresponded to the P3b and posterior positivitycomponents was not significantly greater than zero, r = −.008,p = .973, despite our use of a Promax rotation, which allowsfactors to be correlated. Future research should be directed atexploring experimental manipulations that might dissociatethese two components.

Conclusions

We explored the neural signature of EIB, a phenomenonwhereby emotional distractors impair the detection of targetsthat appear soon after them. We found that negative emotionaldistractors elicit a large N2 component (perhaps reflectingautomatic capture of attention) and a P3b component (thoughtto reflect consolidation of information into WM). The ampli-tudes of these components were inversely related to the am-plitude of the same components elicited by the target, suchthat targets presented soon after emotional distractors showedsmall N2s and small P3bs. In addition, the likelihood of errorsin reporting the target in the EIB task was related to the

magnitude and timing of the N2 and P3b components elicitedby the negative distractor. Emotional distractors also elicited aposterior positivity resembling the Pd component, whichmight reflect the process of disengaging attention from thenegative picture.

Are the same mechanisms that are at work in the AB alsoresponsible for EIB? We found some striking similaritiesbetween ERP effects observed in the two paradigms. In bothcases, an inability to report a target following an earlier salientinput was associated with a suppression of the N2 and P3bcomponents associated with the target. Our results show thatemotional distractors capture attention early enough in pro-cessing to be reflected in the N2 component and that theydisrupt attentional allocation to, and WM consolidation of,subsequently presented targets. EIB and the AB appear toinvolve the same limited-capacity mechanism that serves asa bottleneck in the processing of rapidly presented visualinformation. The question as to whether this overlap is com-plete, or whether there may yet be some unique aspect to theroute through which emotional distractors disrupt target per-ception, awaits further investigation.

Acknowledgments This work was supported by NIH grant R03MH091526 to S.B.M. and NSF grant BCS-1059560 to J.E.H. We wishto thank Matthew Doran, Scott McLean, and Lingling Wang for helpfuldiscussions about this project.

References

Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10,433–436.

Cardinal, R. N., & Aitken, M. R. F. (2006). ANOVA for the behavioralsciences researcher. Mahwah: Erlbaum.

Chun, M. M., & Potter, M. C. (1995). A two-stage model for multipletarget detection in rapid serial visual presentation. Journal ofExperimental Psychology. Human Perception and Performance,21(1), 109–127.

Dell’Acqua, R., Jolicoeur, P., Luria, R., & Pluchino, P. (2009).Reevaluating encoding-capacity limitations as a cause of the atten-tional blink. Journal of Experimental Psychology. HumanPerception and Performance, 35(2), 338–351.

Dien, J. (2010). The ERP PCA Toolkit: An open source program foradvanced statistical analysis of event-related potential data. Journalof Neuroscience Methods, 187(1), 138–145.

Dux, P. E., & Marois, R. (2009). The attentional blink: A review of dataand theory. Attention, Perception, & Psychophysics, 71(8), 1683–1700.

Dux, P. E., Asplund, C. L., & Marois, R. (2009). Both exogenous andendogenous target salience manipulations support resource deple-tion accounts of the attentional blink: A reply to Olivers et al.Psychonomic Bulletin and Review, 16(1), 219–224.

Flaisch, T., Junghöfer, M., Bradley, M. M., Schupp, H. T., & Lang, P. J.(2008). Rapid picture processing: Affective primes and targets.Psychophysiology, 45(1), 1–10.

Foti, D., Hajcak, G., &Dien, J. (2009). Differentiating neural responses toemotional pictures: evidence from temporal-spatial PCA.Psychophysiology, 46(3), 521–530.

Cogn Affect Behav Neurosci

Keil, A., Debener, S., Gratton, G., Junghöfer, M., Kappenman, E. S., Luck,S. J.,⋯ Yee, C. M. (2014). Committee report: Publication guidelinesand recommendations for studies using electroencephalography andmagnetoencephalography. Psychophysiology, 51(1), 1-21.

Kiss, M., Grubert, A., Petersen, A., & Eimer, M. (2012). Attentionalcapture by salient distractors during visual search is determined bytemporal task demands. Journal of Cognitive Neuroscience, 24(3),749–759.

Kranczioch, C., Debener, S., Maye, A., & Engel, A. K. (2007). Temporaldynamics of access to consciousness in the attentional blink.NeuroImage, 37(3), 947–955.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2001). InternationalAffective Picture System (IAPS): Instruction manual and affectiveratings (Tech. Rep. No. A-5). Gainesville: University of Florida,Center for Research in Psychophysiology.

Luck, S. J. (2011). Electrophysiological Correlates of the Focusing ofAttention within Complex Visual Scenes: N2pc and RelatedElectrophysiological Correlates. In S. J. Luck & E. Kappenman(Eds.), Oxford handbook of event-related potential components.New York: Oxford University Press.

Luck, S. J., & Hillyard, S. A. (1994). Spatial filtering during visualsearch: Evidence from human electrophysiology. Journal ofExperimental Psychology. Human Perception and Performance,20(5), 1000–1014.

Luck, S. J.,Woodman, G. F., &Vogel, E. K. (2000). Event-related potentialstudies of attention. Trends in Cognitive Sciences, 4(11), 432–440.

Martens, S., &Wyble, B. (2010). The attentional blink: Past, present, andfuture of a blind spot in perceptual awareness. Neuroscience &Biobehavioral Reviews, 34(6), 947–957.

Most, S. B., Chun, M. M., Widders, D. M., & Zald, D. H. (2005).Attentional rubbernecking: Cognitive control and personality inemotion-induced blindness. Psychonomic Bulletin and Review, 12,654–661.

Olivers, C. N. L., van der Stigchel, S., & Hulleman, J. (2007). Spreadingthe sparing: Against a limited-capacity account of the attentionalblink. Psychological Research, 71, 126–139.

Pelli, D. G. (1997). TheVideoToolbox software for visual psychophysics:Transforming numbers into movies. Spatial Vision, 10, 437–442.

Peyk, P., Schupp, H. T., Keil, A., Elbert, T., & Junghöfer, M. (2009).Parallel processing of affective visual stimuli. Psychophysiology,46(1), 200–208.

Picton, T. W., Bentin, S., Berg, P., Donchin, E., Hillyard, S. A., Johnson,R., ⋯ Taylor, M. J. (2000). Guidelines for using human event-related potentials to study cognition: Recording standards and pub-lication criteria. Psychophys, 37(2), 127-152.

Polich, J. (2012). Neuropsychology of P300. In S. J. Luck & E. S.Kappenman (Eds.), Handbook of event-related potentialcomponents (pp. 159–188). New York: Oxford University Press.

Reiss, J. E., Hoffman, J. E., Heyward, F. D., Doran, M. M., &Most, S. B.(2008). ERP evidence for temporary loss of control during the

attentional blink. Naples: Poster presented at the annual meetingof the Vision Sciences Society.

Sawaki, R., & Luck, S. J. (2010). Capture versus suppression of attentionby salient singletons: electrophysiological evidence for an automaticattend-to-me signal. Attention, Perception, & Psychophysics, 72(6),1455–1470.

Sawaki, R., Geng, J. J., & Luck, S. J. (2012). A common neural mech-anism for preventing and terminating the allocation of attention. TheJournal of Neuroscience, 32(31), 10725–10736.

Schupp, H. T., Junghöfer, M., Weike, A. I., & Hamm, A. O. (2003a).Attention and emotion: An ERP analysis of facilitated emotionalstimulus processing.Cognitive Neuroscience and Neuropsychology,14(8), 1107–1110.

Schupp, H. T., Junghöfer, M., Weike, A. I., & Hamm, A. O. (2003b).Emotional facilitation of sensory processing in the visual cortex.Psychological Science, 14(1), 7–13. Elsevier.

Schupp, H.T., Flaisch, T., Stockburger, J., & Junghöfer, M. (2006).Emotion and attention: Event-related brain potential studies. InAnders, Ende, Junghöfer, Kissler, & Wildgruber (Eds.), ProgBrain Res, Vol. 156 pp. 31-51.

Sergent, C., Baillet, S., & Dehaene, S. (2005). Timing of the brain eventsunderlying access to consciousness during the attentional blink.Nature Neuroscience, 8(10), 1391–1400.

Shapiro, K., Schmitz, F., Martens, S., Hommel, B., & Schnitzler, A.(2006). Resource sharing in the attentional blink. NeuroReport,17(2), 163–166.

Sheppard, D. M., Duncan, J., Shapiro, K. L., & Hillstrom, A. P. (2002).Objects and events in the attentional blink. Psychological Science,13, 410–415.

Spencer, K. M., Dien, J., & Donchin, E. (2001). Spatiotemporal Analysisof the Late ERP Responses to Deviant Stimuli. Psychophysiology,38, 343–358.

Vogel, E. K., & Luck, S. J. (2002). Delayed working memory consolida-tion during the attentional blink. Psychonomic Bulletin and Review,9(4), 739–743.

Vogel, E. K., Luck, S. J., & Shapiro, K. L. (1998). Electrophysiologicalevidence for postperceptual suppression during the attentional blink.Journal of Experimental Psychology. Human Perception andPerformance, 24(6), 1656–1674.

Wang, L., Kennedy, B. L., &Most, S. B. (2012).When emotion blinds: Aspatiotemporal competition account of emotion-induced blindness.Frontiers Psychology: Special Topic on Emotion and Cognition, 3,438.

Weinberg, A., & Hajcak, G. (2011). The late positive potential predictssubsequent interference with target processing. Journal of CognitiveNeuroscience, 23, 2994–3007.

Woodman, G. F., Arita, J. T., & Luck, S. J. (2009). A cuing studyof the N2pc component: An index of attentional deploymentto objects rather than spatial locations. Brain Research, 1297,101–111.

Cogn Affect Behav Neurosci